A new real-time 3D sensor is presented that is capable of delivering 25 high-resolution depth maps per second with a representative measurement error of 0.2 mm. The sensor is well suited for real-world application as it imposes no major scene constraints, i.e. it copes well with colored objects, moving scenes and background illumination. The acquisition technique, which is a form of structured light with a single projection pattern, is introduced, where the focus is on two aspects: Firstly, a new type of projection pattern that employs local color edge patterns for encoding so that it is possible to decode them even under noisy real-world conditions. Secondly, an algorithm that exploits the pattern properties to recognize the projected color edge patterns reliably. It then describes a 3D sensor based on this technique and presents experimental results obtained with it. It furthermore discusses an exemplary commercial application of the 3D sensor to the task of wheel alignment.